Hello. I am trying to integrate transformers’ wav2vec2 model with hidet but getting a BackendCompilerFailed exception.
The repository with code to reproduce the exception: GitHub - egorsmkv/wav2vec2-hidet
Logs:
- wav2vec2-hidet/logs_with_search_space_2.txt at main · egorsmkv/wav2vec2-hidet · GitHub
- wav2vec2-hidet/logs.txt at main · egorsmkv/wav2vec2-hidet · GitHub
The exception:
torch._dynamo.exc.BackendCompilerFailed: hidet_backend raised RuntimeError: hidet: Conv1d(1024, 1024, kernel_size=(128,), stride=(1,), padding=(64,), groups=16) has no parameter/buffer weight, occurred when interpreting self_conv with
HidetConv1d(tensor(...))
I’d love to hear any feedback on the problem.